Non-specific objects in the pseudopassive : the syntax and semantics of English pseudo-incorporated pseudopassives

(1)

NON-SPECIFIC OBJECTS IN THE PSEUDOPASSIVE

The Syntax and Semantics of English Pseudo-Incorporated Pseudopassives by

JILLIAN LOUISE MILLS

B.A. with Honours, Linguistics, McGill University 2005 Submitted to the Department of Linguistics and Philosophy in partial

fulfillment of the requirements for the degree of Master of Science in Linguistics

at the

Massachusetts Institute of Technology

MASSACHULSETTS INSTITUTE OF TECHNOLOGY

OCT 1

4 2008

L

LIBRARIES

September, 2008

DepartIlent of Linguistics and Philosophy, September 3ra₂₀₀₈

Certified by_

Norvin Richards Associate Professor of Linguistics Thesis Supervisor Accepted by

Irene Heim Professor of Linguistics Head, Department of Linguistics and Philosophy

ACKNOWLEDGMENTS

Boy afraid, prudence never pays, and everything she wants costs money...

'Girl Afraid' -Morrissey (The Smiths)

Being such an important and delicate matter, acknowledgments I imagine should not be left until 1:43am to be written. Unfortunately it's a habit I can't seem to shake.

I wish to thank first and foremost my committee members for getting me through the Master's process, and putting up with my indecisive last minute career changes. Norvin Richards certainly inspired confidence in working on the English dialect I know best when he remarked (upon hearing the pseudopassive sentences I was thinking of studying) that I had a parser of steel. My meetings with Michel Degraff brought up so many more interesting questions and puzzles than I can address in this modest thesis, and he always managed to send me off again with a renewed and improved fascination for my topic. Without Kai von Fintel's scheduling and supportive but firm nudges about deadlines and general awareness of the passing of time, this paper would quite possibly never have ended. And it needed very badly to end and defend, soon.

This paper was influenced by many other MIT faculty members also, as it came about initially piece by piece as squibs for various courses, beginning with a paper on pseudopassives for my Advanced Syntax class with David Pesetsky and Shigeru Miyagawa, then another (adding in the direct objects) for Topics in Syntax (on DP/NP Syntax) taught by David Pesetsky and Sabine latridou, and yet another (focusing on the implications of PIPPs for Burzio's Generalization) for Sabine Iatridou and Elena Anagnostopoulou's More Advanced Syntax course. For each paper I received such helpful and encouraging comments that the topic continued to build and now turned into a Master's thesis I have still barely scratched the surface. Many thanks to these professors in particular, but also to the many others I've had the opportunity to learn from, discuss with, and cat/house-sit for over the past three years.

I had the good fortune to present parts of this study at the ECO5 student colloquium at UMass-Amherst and at the Canadian Linguistics Association meeting, both in 2007; many thanks to the audiences at both of these presentations and their wonderful comments and feedback (and to Omer Preminger for suggesting the PIPP acronym). My ling-05 classmates are certainly due a large thank you for enduring more presentations of this material than anyone else, particularly in our Syntax-Semantics workshop fall 2007. David Hill in particular, as housemate and fellow PIPP-speaker, has positively influenced this paper, my thoughts about Linguistics and human nature in general, and my time here in Cambridge/Somerville to a degree that I don't have words for. Baked goods and pifiatas seem the only appropriate expression of my appreciation.

I've had the benefit of meeting and interacting with so many other talented students at MIT as well, notably Pritty Patel and Patrick Grosz who fed me so well on so many occasions, my office mates (past and present) who were entertaining even when very noisy, fellow MITWPL employees, and everyone else just coming in, leaving, visiting, or otherwise.

Lastly, a thank you to my family who are always supportive no matter how geographically distant, and to the Social Sciences and Humanities Research Council for their financial support (award # 752-2007-0384).

(6)

(7)

TABLE OF CONTENTS

ABSTRACT ... ... ... 3..3

ACKNOWLEDGMENTS ... 5

TABLE OF CONTENTS ... 7

INTRODUCTION ... ... CHAPTER 1 - PROBING THE PROPERTIES OF PIPPS ... 13

1.1 THE GRAMMATICAL STATUS OF PIPPS ... 13

1.2 PIPP-OBJECT PROPERTIES ... 18

1.2.1 In Search of D: No Determiners, No Pronouns, No Possessors... ... 18

1.2.2 Bigger than N ... 19

1.2.3 Adjective Interpretations ... ... 19

1.2.4 Extraction ... 23

1.2.5 Scopelessness ... 24

CHAPTER 2 - THE SYNTAX... ... 27

2.1 ...OF (PSEUDO-INCORPORATED) (PSEUDO)PASSIVES ... ... 27

2.1.1 Passive 101 ... ... 27

2.1.2 Pesetsky & Torrego, and Putting the Passive Pieces Together ... 28

2.1.3 'Reanalysis' Reanalyzed ... 31

2.2 ...OF NON-INTERVENERS ... 35

2.2.1 Closest and Highest... 35

CHAPTER 3 - THE SEMANTICS... ... ... 39

3.1 ...OF PSEUDO-INCORPORATION ... ... ... 39

3.1.1 Cross-Linguistic Support... 39

3.1.2 Combining Predicates with Properties ... ... .41

3.2 ...OF NON-SPECIFIC OBJECTS IN OTHER CONSTRUCTIONS ... ... 44

3.2.1 Existential-There... 44

3.2.2 Intensional Verbs and Opacity... 47

CHAPTER 4 - CONCLUSIONS ... ... 53

4.1 SUMMARY ... 53

4.2 REMAINING QUESTIONS ... ... 53

(8)

(9)

INTRODUCTION

As indicated by its title, this thesis investigates the properties of non-specific objects as they occur in English pseudopassive sentences, which will be shown to, in fact, be the only kind of objects allowed in the pseudopassive construction. To illustrate, the pseudopassive sentence in (1) allows the indefinite object a habanero to occur between the verb and stranded preposition, while the same object preceded by the definite determiner, the habanero in (lb), is judged ungrammatical. Furthermore, a habanero cannot refer to a specific habanero (nor take scope outside of other operators in the sentence, as will be shown later in chapter 1).

(1) a. This cutting-board was chopped a habanero on. b. *This cutting-board was chopped the habanero on.

The theoretical importance of this observation is two-fold. First, it tells us something very interesting about the syntax of interveners (that is, nominals/phrases that block structurally lower phrases from raising over them), or rather the syntax of non-interveners: to block raising-to-subject it is not enough just to be a nominal in an argument position, you also have to have D-level structure. I will argue that those nominals that occur without the required structure and its associated features cannot raise to subject because they are not suitable goals for T's probing -an avocado in (la) just doesn't have what T needs. This frees-up T to look elsewhere to satisfy its phi-feature checking needs, namely into the PP below. In essence, the sentences under study are evidence that nominals exist that in many ways look like other argument DPs - for example they can be conjoined, and modified - but in many other ways behave much differently - they cannot be extracted by A'-movement, cannot take wide scope over other operators, and cannot be modified by relative clauses.

Second, the fact that objects in the pseudopassive are restricted to these reduced, sub-DP, non-specific forms makes this particular construction an optimal research tool for probing the relationship between the syntactic structure and the semantic interpretation of nouns. As we will see shortly, the missing D-level projections of these nominals correspond to missing semantics, most notably specificity/definiteness, and wide-scope readings, but also certain interpretations of otherwise ambiguous modifiers are not possible. This thesis serves, in a sense, as a first-pass probing into these syntax-semantics correlations, and as such will at times raise more questions than I can answer within its limits.

More generally the availability of any object in sentences like (1), for at least some English speakers, is unexpected under previous analyses of the pseudopassive construction (namely that of Hornstein & Weinberg 1981). Since these theories have posited an adjacency requirement on the verb and preposition (in order for them to form a complex predicate by reanalysis), allowing an object between the two shouldn't be possible. A new approach to the mechanics of the common pseudopassive is therefore in order, and I propose that the current theoretical framework of Pesetsky & Torrego (2004, 2006, 2007) can be smoothly applied to analyze the pseudopassive as an instance of Tense-sharing between T, the passive predicate, and the stranded preposition.

The topic of specific objects has seen a recent flurry of research, mostly from non-English language data where these objects are more easily distinguished from other indefinites by their special syntax, semantics, and morphology (Bittner 1984,1987, van Geenhoven 1998,

(10)

Wharram 2003, Massam 2001, Dayal 2003, Coon to appear, Chung & Ladusaw 2004, Deal 2007, and others). We will see below that the curious properties observed for objects in the pseudopassive pattern with many (if not most) of the ones reported in those studies, which cut across languages and language families from Mayan to Austronesian to Eskimo-Aleut and beyond. This thesis also, then, adds to the growing literature and deeper understanding of the typology of these nominal objects.

What I will come to defend herein is that, following many of the authors just cited, pseudopassive objects are interpreted as properties of type <et>, and not as individuals. This property-type semantics corresponds to the lacking D-level structure in the syntax, and together these features account for their distinctive properties, namely the ungrammaticality of strong determiners, pronouns, or possessives, their scopelessness, inability to extract or to be modified by relative clauses, and their necessarily non-specific interpretation. In many languages these objects appear intractably adjacent to the predicate similar to lexical noun-incorporations like to cat-sit, or apartment-hunt, earning them the term pseudo-incorporated, and the semantic mechanism by which the two come to be joined in meaning pseudo-inco7poration (also known as semantic incorporation). Anticipating this analysis, I will refer to the construction in (1) as

pseudo-incorporated pseudopassive, or PIPP for short.

The obvious question that this paper will not answer conclusively is that of what separates speakers who accept (la) and other PIPPs productively from so many other speakers who are much less permissive with this construction. This appears to be a complicated matter which I will leave for the most part for future research. Chapter 1, however, opens with a discussion on the grammatical status of PIPPs and a number of examples, forms and settings that non-PIPP speakers may find more palatable, from Bolinger (1977). The chapter then proceeds to lay out a number of properties that support the lack of D-level syntax in PIPP direct objects, or that our theory will need to account for otherwise.

In chapter 2, the syntactic structure that I posit for PIPPs will be presented, building up from the essential pieces: the syntax of passives, then of pseudopassives, and finally of pseudo-incorporated pseudopassives. Here I will discuss the issue of interveners, why we might expect the direct object in PIPPs to be one, and how its truncated D-less structure allows it not be. This chapter ends with the syntactic structure that the semantics of the third chapter will be handed to deal with, most notably a sentence with a predicate-type object in place of an individual one.

The third chapter reviews the properties documented for pseudo-incorporated objects cross-linguistically. This serves both as support that we are on the right track in identifying PIPP direct objects as deserving of the title, and to introduce the semantic mechanisms those researchers have proposed to get predicates to take properties as arguments. The proposed mechanisms span from lexical rules/operations (Dayal 2003, van Geenhoven 1998), to new rules of semantic composition (Chung & Ladusaw 2004), to silent morphemes that attach to verb roots (Wharram 2003, Deal 2007). Although empirical data in support of any of these approaches over the others is currently weak, Wharram 2003 argues that the antipassive morpheme in West Greenlandic correlates with the semantically-incorporated behaviour of nominals. He thus attributes this morpheme the semantic function of turning individual-taking verbs into property-taking ones; this morpheme, ANTIP, is simply silent in many other languages. I follow Wharram (and Deal 2007), therefore, in positing a silent morphological head in PIPPs that allows the verb to combine with the D-less, truncated nominals under investigation. Although it is tempting to suggest that the availability of this morpheme is what separates PIPP speakers from non-PIPP speakers, chapter 3 concludes with a discussion of two other cases in English where nominals

(11)

have been argued to semantically incorporate with their predicates: in existential-there constructions, and as complements to intensional verbs on their opaque readings. Chapter 4 will conclude.

To summarize a last time before getting on our way, this thesis argues the following main points:

i. The direct object in PIPPs is necessarily syntactically reduced, missing D-level structure, at the very least.

ii. Without D-level structure, PIPP direct objects do not act as interveners to movement of lower phrases (allowing the lower PP-object to raise to subject).

iii. Pseudopassivization arises by the selection of PP whose P-head is unvalued for T (tense) features by a verb who is also unvalued for these features (i.e., a passive verb); both heads therefore depend on a higher, T-valued head to value their features, creating a complex predicate.

iv. Nominals missing D-level structure syntactically are interpreted semantically as properties (<et>).

v. PIPP predicates combine with a silent head ANTIP yielding a function that takes a property <et> argument, as required to combine with its complement.

(12)

(13)

CHAPTER

1

- PROBING THE PROPERTIES OF PIPPS

1.1 THE GRAMMATICAL STATUS OF PIPPS

Before setting forth on this exploration of the syntactic and semantic properties of PIPPs, some ink must be spilled on the subject of their grammatical status in English. As mentioned above, many of the sentences that have been introduced and will be discussed throughout this study are not accepted out of the blue by most English speakers consulted, or not without difficulty and wincing. There are two points to consider, however, that may help the reader suspend their disbelief that PIPPs could be a part of any English speaker's grammar.

First, PIPPs that are constructed from idioms of the form V+Obj+P seem to be acceptable to all speakers, an observation found in much of the Linguistics literature on (common) pseudopassives also (Jespersen 1928, Mincoff 1958, Hudson 1967, Labov 1972, Bolinger 1977, Hornstein & Weinburg 1981, Baltin & Postal 1996, Tseng 2007). This includes such strings as make fun of take umbrage at, take advantage of make a fighting ground of make allowances for, keep tabs on, etc.; examples are readily observed in modem dialogue from pop-culture

television to e-bay advertisements:

(2) "I've been reading up on your poor nurse's condition, and it is nothing to be made fun of"

South Park, Season 2 'Conjoined Fetus Lady' Episode (3) "Sega Genesis system has been used and taken good care of"

E-bay advertisement for Sega Genesis system (4) "Do I look like someone to be made a jerk of?"

American Dad The structure is therefore not altogether foreign to most English speakers, despite that it does not appear to be as productive for everyone as it is for the speakers of the dialect under study.

Still, there are contexts beyond idioms even that improve the interpretability of PIPPs, as Bolinger (1977) discusses in great detail. Bolinger takes a strong stance on the grammatical productivity of PIPPs and common pseudopassives (referred to as transitive and ditransitive prepositional verbs in his work), maintaining that, "while a running frequency count might well

show that the great majority of prepositional verbs are lexicalized, lexicalization is only contributory. Passive prepositional verbs are a completely open set" (Bolinger 1977: 59). Some forms may seem odd used out of the blue, but put in the proper context they become more natural - this much seems true, although we will see that many of Bolinger's explanations for this are unclear and out of date. Hopefully future work will be able to scrutinize the following examples/contexts and extract the underlying mechanisms at play.

Bolinger argues that two factors regulate the acceptability of passivizing a prepositional verb: the semantic relationship between the prepositional verb and the patient, and the interpretability of the sentence. The first point relates to the well-known requirement that the subject of a passive must be understood as somehow affected by the event described by the predicate. Taking the example (1) from above of cutting-boards and habanero, my chopping a

(14)

habanero on the cutting-board likely got spicy oils and seeds from the pepper all over the board (rendering it unsuitable to use when we cut, say, the coffee cake) - thus the board is interpretable (to some) as having been affected by the cutting of a habanero on it and can serve as subject of the formed passive.

The second criterion relates more or less to sentence processing. As Bolinger explains: we reject a sentence like *I don't like to be blamed mistakes on in spite of the acceptability of the corresponding active Don't blame those mistakes on me and of similar passives like I don 't like to be thrown mud at because of the high degree of expectation that blame will have a personal and not an abstract object: the hearer gets as far as blamed and is sure that I don't like to be blamed means what it appears to mean - and then mistakes on forces him to revise. Even though theoretically the same cumulative ambiguity afflicts That argument was built a case against (first, That argument was built; second, That argument was built a case), the distracting interpretations are less insistent (Bolinger 59).

Again this second component is vague and quite difficult to support without carrying-out proper processing experiments. His point is that the contexts he identifies as facilitating "help the listener to quickly identify the prepositional verb as a unit" (60). I present each of these contexts in turn with some of his examples:

i. Contextual parallelism

(5) a. The defendants - the ones arrested and brought charges against yesterday - are all expected to plead innocent.

b. The books that were separated and stuck markers in are on the middle shelf.

c. This tool has never been used for its main purpose - in fact, it's never been done anything with at all.

In (5), by conjoining or juxtaposing the PIPP with another, more pedestrian passive predicate, the former is understood more quickly as a unit on par with the latter by the listener.1

ii. Analogy

(6) _{a. Mary was written a letter to. (on analogy to Mar; was written a letter)} b. John was done a favor for. (vs. John was done a favor)

c. That's a problem that needs to be found a remedy for. (vs. ...to be found a remedy) The idea here seems to be that PIPPs created from double object verbs can be more easily passivized on analogy to their regular double object parallel, the preposition left stranded at the end adding only some minor extra semantic specification of direction.2

Norvin Richards (p.c.) suggests one possible explanation for this particular helping context: by conjoining a regular passive (or even active) VP with the PIPP one the sentence is allowed by the Principle of Minimal Compliance (Richards 1998). Despite that raising the PP-object from the PIPP VP might be ungrammatical, since it is also raising from a grammatical site within the regular passive VP the two VPs taken together comply minimally with proper movement rules.

2 An interesting though not yet understood detail to note here is that for some PIPP speakers, certain double object

predicates that typically do not work as well as dative (NP-PP) constructions improve when passivized. These speakers judge (id) better than (ic), although not on par with (ia) or its passive (ib):

(15)

iii. The use of empty verbs

(7) a. If these bridges are put supports under, they will withstand any amount of flooding. b. That product can't be made a profit from.

c. What's he miffed about? Was he taken a swipe at by somebody?

Bolinger indicates that the use of empty verbs such as make, take, and put in (7) tend to suggest that a noun will naturally follow them. The V-NP-P string typically has another single word predicate synonymous in meaning, so to put supports under = to support, to make a profit from =

to profit, to take a swipe at = to swipe (at); that is, these verbs are "in the nature of compounding

elements."

iv. Predictability of the verb + noun combination (8) a. John was held a grudge against.

b. The troops were opened fire on.

c. That whole platoon was simply made mincemeat of.

Here, the more predictable or frequent the verb+noun combination (or the

verb+noun+preposition), the more natural the PIPP. Interestingly, as we vary the verb or object to a less common substitute an originally well-constructed PIPP becomes more difficult to accept, shown in the contrasts between (9a) versus (9b), and (9c) versus (9d):

(9) a. I don't like to be brought charges against. b. ?I don't like to be lodged charges against.

c. The solution has been raised serious doubts about. d. ?The solution has been raised serious questions about.

Next, Bolinger observes that using a verb-object-preposition string in the active first facilitates interpretation of the same string in the passive later on in the discourse. Examples are in (10):

v. Repetition and presuppostion

(10) a. -He paid too much for his coat.

-Well, lots of things are paid too much for these days. You have to expect it. b. -What's that pipe for?

-To blow bubbles with. -Well, has it been? -Been what?

-Blown any bubbles with yet?

Similar to the first factor, contextual parallelism, by hearing the complex predicate string first in a typical setting (the active, without any passive-movement), the hearer is better prepared to hear and interpret that same string as a predicative unit the second time around in the passive.

b. Amy was given a headache. c. *Igor gave a headache to Amy. d. ?Amy was given a headache to.

(16)

Constructions that presuppose a prior context for the complex predicate, without the explicit repetition shown above, also have an optimizing effect:

(11) a. *Were you aimed a gun at?

b. How does it feel to be aimed a gun at? c. *This bed has been eaten potato chips in.

d. To be eaten potato chips in isn't the best thing for a bed.

As Bolinger explains, (1 ib) (vs. (1 la)) "contains a petitio principii: it cannot be asked unless the interlocutor has already had a gun aimed at him" (63). He believes that topicalized infinitives such as (11d) are preferred to (11 c) for similar reasons, namely they imply that the matter has been brought up already. Bolinger takes these five situations to support his claim that prepositional verbs exist beyond lexicalized forms, that "the only real restrictions are clarity and intent. The use of a prepositional verb has to be motivated."

We learn from Bolinger's discussion that the grammatical status of PIPPs in standard English is not a simple matter, but more importantly, that evidence of their availability in the right context (and with a varying degree of productivity) is easier to come by than seems from the outset. For the purposes of this paper, however, we limit our investigation to the grammars of those speakers I have found who readily accept PIPPs out of the blue with considerable productivity. That is, if we consider PIPP-acceptability as a continuum across the grammars of different speakers (which seems fairly accurate from my informal surveying of various speakers and dialects), with the speakers who only allow fully lexicalized forms at one extreme, and those who accept a range of forms, provided they occur in the more propitiatory contexts discussed above, fall at various points in the middle, then the speakers under study herein sit at the other extreme of the continuum (I will refer to those at the PIPP-accepting extreme as PIPP-speakers throughout, to facilitate discussion).

From my informal survey of native English speakers (mostly Linguistics students and their friends/family, by e-mail and in person), PIPP-acceptance does not appear to correlate strongly with regional dialect, although most of the PIPP-speakers I found grew up in various locations in Canada (it is worth noting that Dwight Bolinger, however, did not grow up in Canada). Even speakers within the same family are often split with regards to this construction. Judgments throughout are primarily my own, with secondary verification and consultation from other speakers. Curiously, even those who do not fully accept PIPPs out of the blue still share the judgments that definite direct objects are worse than indefinites, and other key PIPP properties/restrictions that I come to shortly.

Lastly, a few words on the status and importance of PIPPs cross-linguistically. In section 2, I present evidence that the direct object in PIPPs is particularly restricted to non-specific indefinite NPs, which have been shown for a wide range of languages to differ syntactically and semantically from other, fuller DPs. PIPPs, therefore, add further support for the existence of this distinction between different kinds of nominal arguments.

Whether other languages might allow a PIPP construction, that is, a similar situation of passive-raising an argument from a PP, over another nominal, and stranding the preposition itself, is an interesting question also. We expect the construction to be very rare indeed, given the unpopularity of one of its necessary parts, preposition-stranding; worse yet, only a subset of languages allowing preposition-stranding by A'-movement also allow the A-movement stranding that yields pseudopassives. Languages known so far to allow pseudopassives include English,

(17)

Norwegian, Vata, and Gbadi (Abels 2003 and references therein), and so only in these languages can we hope to find other PIPP-like constructions. Although investigation and comparison of cross-linguistic PIPP constructions and properties is not the focus of the present paper, I note in passing that Norwegian does appear to allow very PIPP-like sentences indeed, as shown in (12): (12) a. at breveti

that letter-the 'the letter was (Taraldsen 1979, Afarli

ble klisteretfrimerker pd ei. was pasted stamps on pasted stamps on.' 1992)

b. at frimerkeri ble klisteret ei pd brevet. that stamps was pasted on the letter 'the stamps were pasted on the letter' c. at det ble klisteret frimerker pd brevet.

that it was pasted stamps on the letter lit. 'it was pasted stamps on the letter' (Haider 46)

Just as in the English gloss, the subject of (12a) has been raised out of the PP despite the presence of a direct object, frimerker 'stamps.' However unlike English, Norwegian also allows the same sentence with an expletive subject and no argument movement, known as the 'impersonal passive' in (12c). As a preview of some restrictions on PIPP objects we come to below, these same restrictions hold for the direct object in the Norwegian construction, shown in (13): the definite object frimerkene in (13a) is banned, as is the possessive frimerkene mine in (13b), and the universally quantified alle frimerker in (13c).

(13) a. *at breveti blei

that letter-the was 'the letter was pasted the

b. *at breveti blei

that letter-the was 'the letter was pasted my

c. *at breveti ble

that letter-the was 'the letter was pasted all

klistra frimerkene pasted stamp-the stamp on.' klisteret pasted stamps on.' klisteret pasted stamps on.' pd ei. on frimerkene stamps-the mine my.pl alle frimerker all stamps

(Sverre Johnsen, personal communication)

Interestingly, the definiteness restriction on the direct object holds for the impersonal forms of these sentences also. Further investigation into the differences and similarities between PIPP constructions in those languages that can create them, although interesting and relevant to the arguments of this paper, must be left for future work.

Having now discussed the status of PIPPs, we move on in the next section to a more detailed look at the properties of this interesting construction.

pc ei. on

pd ei. on

(18)

1.2 PIPP-OBJECT PROPERTIES

1.2.1 In Search of D: No Determiners, No Pronouns, No Possessors...

When we take a closer look at what types of nominal phrases are possible in PIPP direct object position, we observe that items which usually (i.e. are argued to) inhabit D are illicit in the pseudopassive direct object.

First off, pseudopassive direct objects cannot be preceded by strong determiners or quantifiers, seen in (14):

(14) a. *This cutting-board was chopped the avocado on. b. *This cutting-board was chopped every' avocado on.

c. *This cutting-board was chopped some avocados on (...the rest are on the table) d. *This cutting-board was chopped an avocado on (the one that was on the table).

It is important to note that the acceptability of indefinite determiners, as in (14b&c), varies with their interpretation as either strong or weak (i.e., specific vs. non-specific, or put differently, presuppositional vs. non-presuppositional); the sentences in (14c&d) are fine on a weak/existential reading. Crucially, weak interpretations of indefinite determiners have been argued in the literature to derive from their ability to combine with nouns as modifiers, rather than as generalized quantifiers (Partee 1989, Diesing 1992, Alonso-Ovalle & Menendez-Benito 2002). When understood referentially, as indicated in the examples above, the indefinites contain true determiners, i.e. D-heads, and the sentences are ungrammatical.

In (15), we see that pronouns are also not possible:

(15) a. -What about the students? *-The NSF grant was paid them with.

(vs. 'The NSF grant was paid students with')

b. -Where'd this dirt come from? *You were dumped it on, weren't you? (vs. 'you were dumped dirt on')

Finally, possessors, which are argued for English to be in D or SpecDP (Cardinaletti 1998, Larson & Cho 2003, and others), cause the sentences to be ungrammatical:

(16) a. *Those pants just aren't meant to be tucked Blanche's shirts into. b. *These pants just aren't meant to be tucked my shirts into.

c. * This cage was kept Jurassic Park's Raptors in, but they've escaped

Given that those items expected to occur in D are not possible in PIPP direct object position, it is logical to suspect that there is no D projection in these nominal phrases. The next section briefly demonstrates that although the DO is smaller than a DP, it is larger than just an N-head.

(19)

1.2.2 Bigger than N

In contrast to lexical noun-incorporations (baby-sitting, salamander-hunting, etc.), PIPP direct objects appear to be larger phrases supporting modifiers, weak quantifiers (as noted above), and conjunction. We see in the following examples that modifiers of the direct object are readily accepted (and even publicly attested) in the pseudopassive:

(17) a. Sega Genesis system was used and taken good care of. (from a merchandise-description on ebay)

b. This cutting board was chopped various different local vegetables on. c. Wow, this fabric could be made lovely cushions out of.

Conjoined modified NPs are possible also:

(18) a. That cutting-board was only chopped purple onions and fresh okra on. b. This cage is kept velociraptors and baby tyrannosaurs in.

c. The stolen money was bought a sports car and some liquor with. Again, definite/strong quantifiers are not possible in conjunctions either:

(19) a. *That cutting-board was only chopped the onions and the tomatoes on. b. *This cage is kept every velociraptor and (that) baby tyrannosaur in. c. *The stolen money was bought the sports car and that liquor with.

The possibility of modification and conjunction are both strong support that the DO is not a simple N-head. The position and interpretation of modifiers is however restricted in interesting ways, which I elaborate on presently.

1.2.3 Adjective Interpretations

It is well-known since Bolinger (1967) and much work since (Larson 1998, 2001, Cinque 2005, amongst others) that certain adjectives in English can appear both pre- and post-nominally: in pre-nominal position these adjectives are ambiguous between two different interpretations, whereas in post-nominal position the ambiguity is resolved. The most famous contrasts occur between stage- and individual-level, restrictive versus non-restrictive, and implicit relative versus modal readings, demonstrated in (20):

(20) a. Stage level vs. Individual level

visible stars =stars inherently visible (i-level), or stars visible right now (s-level) stars visible =stars visible right now (s-level)

ex. The visible stars are Polaris and Alpha Centauri... ...in this hemisphere.

(i-level: these stars are inherently visible given their brightness and proximity to Earth)

(20)

...all the others are blocked by the clouds. (s-level: these stars are the ones we can see right now)

The stars visible (tonight) are Polaris and Alpha Centauri. (only s-level) b. Restrictive vs. Non-restrictive

every unsuitable word =every word is unsuitable (non-restrictive), or those words that are unsuitable (restrictive)

every word unsuitable =every word that is unsuitable (restrictive) ex. The unsuitable words were cleaned off of the building wall...

...now it's clean as new.

(non-restrictive: the words were all unsuitable) ...the suitable words were left in place. (restrictive: of the words, those that are unsuitable)

The words unsuitable (for displaying on building walls) were cleaned off. (only restrictive)

c. Implicit relative vs. Modal

every possible candidate =every potential candidate (modal), every candidate possible (implicit relative) every candidate possible =every candidate it was possible for us to interview

(implicit relative) ex. They interviewed every possible candidate...

...then asked which ones wanted to contend. (modal)

...then there was no one left to interview and they gave up. (implicit rel.) They interviewed every candidate possible. (only implicit rel.)

Interestingly, when we use these adjectives to modify pseudopassive direct objects, however, only one interpretation is possible, shown in the following examples:

1. Stage-Level vs. Individual-level

(21) a. Here, this table is to be noted visible stars in for each hour of your observation, #(and whether you could see them or not).)

a'. You should note visible stars in this table for each hour of your observation. b. *Where's an astronomy book that's listed stars visible in?

c. Where's an astronomy book that's listed visible stars in? c'. The authors listed visible stars in this astronomy book.

The hash-mark outside of the brackets indicates that the utterance would be awkward, and rather inappropriate

without the bracketed continuation.

(21)

d. #This fancy telescope was to be counted visible stars with, but the clouds were so thick that there weren't any.

d'. I was going to count visible stars with this telescope, but there weren't any. e. Did you hear about that botched-up burglary back in the 50's? Well there's the

prison that was locked-up responsible individuals in... #(until they were finally acquitted years later).

e'. They locked up responsible individuals in that prison.

In (21a), the pragmatically preferred context would be one in which the stars that are actually visible at each hour of observation (given cloud-cover, light-pollution, etc.) are the ones being noted in the table; the bracketed continuation disambiguates the context to one that instead forces the individual-level interpretation of 'visible stars.' That (21a) is awkward without this continuation indicates that the stage-level interpretation is inaccessible - the only reading available is therefore the individual-level one which clashes with the preferred context until the continuation is parsed. Similarly, in (21 d) the sentence is interpreted as a contradiction since our world knowledge maintains that the set of inherently visible stars is constant (modulo the occasional super-nova), while the sentence asserts that there weren't any such stars. This contradiction is unexpected if the pre-nominal adjective 'visible' is ambiguous between stage-level and individual-stage-level readings.

The post-nominal position for these adjectives is completely ungrammatical, shown in (21b). The example in (21e) is similar to (21a) - responsible individuals in this sentence can only be understood as 'people who are generally responsible individuals,' and not as 'people who were responsible for the crime.' The interesting observation from (21) is that only the individual-level interpretations of the modifiers in question are possible.

II. Restrictive vs. Non-restrictive

(22) a. I'll never hire that baby-sitter again - my children were yelled several

unacceptable curse-words at while we were out... (#we only yell the acceptable ones at them).

a'. The baby-sitter yelled several unacceptable curse-words at our child. (we only yell the acceptable ones at them).

b. *My children were yelled several curses (that were) unacceptable at. c. Igor's editorial was deleted tenable accusations from... (#of course, the

untenable ones were left in).

c'. Igor deleted tenable accusations from his editorial... (of course the untenable ones were left in).

Again in (22a,c) we observe that only one reading, the non-restrictive one, is possible for the typically ambiguous (as in (22a', c') modifiers unacceptable and untenable. (Example (22b) shows that these adjectives are not possible in post-nominal position either.)

II. Implicit Relative vs. Modal

(23) a. This computer program will be background-searched possible candidates with... (# then we'll sort out which ones have potential).

(22)

a'. We will background-search possible candidates with this program... then we'll sort out which ones have potential.

b. *This program will be background-searched candidates possible with.

Lastly, possible in (23a) can only mean 'potential candidates,' its modal reading, and not '(all) candidates that it was possible for us to compile,' which would be the implicit relative reading. The non-PIPP version allows both interpretations.

One way to analyze this data is to follow Larson (1998) in assuming that the ambiguity typically observed for these modifiers falls out from the syntactic level, either NP or DP, that is modified. Stage-level, restrictive, and implicit relative interpretations arise when the adjective modifies the DP-level (a in (24)); individual-level, non-restrictive, and modal interpretations arise when the adjective modifies the NP-level (P in (24)). It's worth noting that only the DP-level modifiers surface both pre- and post-nominally, recalling that post-nominal ones are all illicit in PIPPs.

(24) a. [DP Da a[NP P N] a] (a = DP modifier; P = NP modifier)

(Larson & Marunii 2004: 280)

If the partitioning into DP-level and NP-level modifiers represented in (24) is correct, then the results of the above examples correlate nicely with those of the previous sections. That is, if PIPP direct objects are necessarily NPs, then it is expected that DP-level modification is not possible. The syntactic and semantic structures underlying these modifier ambiguities are in hot debate, however, and there is little agreement whether Larson's or any other proposed structure is correct (see Alexiadou 2001, Cinque 2005, Larson 1998, inter alia).4 Despite their differences, both Alexiadou's (2001) and Cinque's (2005) analyses for ambiguous modifiers are also compatible with the idea that PIPP-objects lack a D-level (and possibly more levels directly below).

Alexiadou posits a dual source for nominal adjectives: direct and indirect modification. Direct modification yields those adjective meanings that are only available pre-nominally in English, and, as the term suggests, combine with the noun directly. Indirect modification is made via relative clauses that may or may not have been reduced, and these are the adjective meanings that are available on either side of the noun. Alexiadou assumes Kayne's (1994) analysis of relative clauses, in which a CP is selected by D, and the head-DP of the relative clause raises from within CP to its specifier. This structure is represented in (25):

(25) a. [DP D CP ]

b. [DP D [CP DPj [ C' [Ip ... tj ... ]]]] head-raising (Alexiadou 2001: 17, from Kayne 1994)

As we see, the D level on this account is again crucial for the stage-level, restrictive, and implicit-relative interpretations, but not for the individual-level, non-restrictive, and modal ones, quite similar to Larson's structure above.

4 The pre-nominal vs. post-nominal modifier picture gets even more complicated when we examine the facts in other languages, as these authors do. It is far beyond the scope of this paper to solve the underlying structure of modifier positions and ordering (although more could be said).

(23)

For Cinque, as for Larson also, adjectives on their individual-level, non-restrictive, and modal interpretations are introduced in the inner-most positions, closer to the noun itself, while they can only receive stage-level, restrictive, and implicit-relative meanings when introduced as relative clauses (or reduced ones) in the outer parts of the nominal phrase. Clearly there must be something, possibly semantic, that forces this syntactic separation into inner and outer layers, although it does not seem to have been identified in the literature as of yet. I suspect given the non-specific semantics of pseudo-incorporated nouns cross-linguistically, that if my argument is sound that PIPP-objects belong to this class, then the outer modifier levels and meanings available there must be involved in introducing specificity and reference to the nominal. This

area of nominal-internal structure is ripe for further study, though I cannot pursue it here.

The modification data presented above, although not entirely conclusive, is explained in a simple manner assuming any of three leading analyses of adjective modification - the solution is that PIPP-objects are missing D-levels (and possibly more just below D). The NP status of the PIPP-object argued for here has other supporting repercussions as well, which I explore in the next sections.

1.2.4 Extraction

A curious fact about PIPP direct objects is that they cannot be extracted overtly by topicalization, in (26a,c), or wh-movement, attempted in (26b,d):

(26) a. *It was habanero that this cutting-board was chopped t on. b. *What was this cutting-board chopped t on?

c. *It was a student the NSF grant was paid t with. c. *Who was the NSF grant paid t with?

The severe ungrammaticality of the attempts in (26) to move these objects in PIPP constructions is in stark contrast to the facility of extracting them from more pedestrian structures, in (27): (27) a. It was habanero that we chopped t on this cutting-board.

b. What do you chop t on this cutting board?

c. It was a student that we paid t with the NSF grant. d. Who did you pay t with the NSF grant?

Cagri (2007) argues that Turkish bare-NPs cannot raise for case, nor undergo wh-movement or topicalization, whereas full DPs can. Thus, PIPP DOs appear again to pattern in

5 Cagri's arguments for this are however too complex to fit in the space of this paper. Minimally, we see in (ib) vs. (ii) that the non-specific object pasta 'cake' occurs without case-morphology, unlike its specific/definite counterpart

bu pasta-yi 'this cake-ACC', and cannot be separated from the predicate by intervening adverbs:

(cont. on next page)

(i) a. Hasan din pasta yedi.

Hasan yesterday cake ate

'Hasan ate (some) cake yesterday.'

b. Hasan pasta *hIzlI /*diin /*kasIkla yedi Hasan cake quickly/yesterday/with a spoon ate

(ii) Hasan bu pasta-yI hIzll /diin /kaslkla yedi

(24)

behaviour with bare-NPs cross-linguistically, and not with DPs. I argue below that extraction of PIPP DOs is blocked by the semantic type of the trace they would leave behind - an individual of type e, and not a property of type <et> which the predicate is anticipating - but we will come to this shortly.

First, however, I wish to explore further the issue of wh-words in PIPPs, for the following reason: wh-words en situ in the PIPP DO position are possible, shown in (28b,c):

(28) a. *How many onions was this cutting board chopped on? b. This cutting board was chopped how many onions on?

c. Quick, I need to know when this cutting-board was chopped what on! (cf. (26b))

Given (28a) and the ungrammatical attempts at extraction above, the acceptability of sentences like (28b) and (28c) is rather surprising. The pattern starts making some sense, at least in light of work on wh-movement and specificity, when we observe that not all wh-words are licensed in this position. If we replace the wh-words in (28b) and (28c) with ones deemed more specific (as in which, who, etc. see Kiss 1993, Munaro & Obenauer 2002), as in (29), the question construction is once again ungrammatical:

(29) a. *This cutting board was chopped which onions on? b. *Which cutting board was chopped which onions on?

c. *Quick, I need to know when this cutting-board was chopped who on? d. *What was paid who with?

A number of researchers have argued that in wh-movement, it is not the wh-word itself that is probed and attracted to C, but a separate head in the nominal left periphery - Q in Cable (2007),

and F (or Foc, for Focus) in Aboh (2004). Based on morphosyntactic data from Gungbe, Aboh (2004) argues that the nominal F head is a realization of the syntactic D-level, which is split into several phrases in parallel to recent split theories of C at the clausal left-periphery. The D-level also introduces specificity, via a nominal TopicP. Our English data is consistent with Aboh's theory if the PIPP-object wh-words lack D-level projections, since this would demonstrate that wh-words are permitted as NP-objects so long as they are non-specific and do not raise: specificity and question raising would require the D-level structure that is argued here to be inadmissible in PIPP DO position.

1.2.5 Scopelessness

Related to the extraction pattern above, PIPP DOs have necessarily narrow scope, unlike their canonical-sentence counterparts. It is well known that multiply quantified phrases in a sentence (in English) generally result in scope ambiguities. In attempting to find scope ambiguities in PIPPs, however, we get the following (some sentences adapted from Carlson 1977, Chierchia

1998) (contrasted again with the non-PIPP (non-bare plural) versions in (30b',c'd',e'):

'Hasan ate this cake (quickly/yesterday/with a spoon). (Cagri 2007:141)

(25)

(30) a. Every cutting-board was chopped vegetables on. b. Each grant was paid three students with.

b'. They paid three students with each grant. c. All vegans have been fed an animal product to...

...by accident at some point in their lives. ...*gelatin was snuck into all water supplies! c'. Someone has fed an animal product to all vegans. d. ?Each lab was tested parts of that machine in. d'. They tested parts of this machine in each lab. e. That cutting-board wasn't chopped a habanero on. e'. I didn't chop a habanero on this cutting-board.

*(some) > every

*three > each

three > each, each > three

all > an

*an > all

all > an, an > all

*(some) > each

parts > each, each > parts

Neg > a, *a > Neg Neg > a, ?a > Neg

Kind-denoting bare plurals such as vegetables in (30a) are scopally inert even in regular contexts, so we do not in fact expect scope ambiguities for them in PIPPs either (which is born-out in (30a)). Necessarily narrow scope would not be expected for the singular indefinites in (30c,e), however, if we thought they were no different from regular DP objects. Non-kind bare plural typically pattern with singular indefinites in terms of scope options, and we would expect the same for PIPP-objects of this kind, but the opposite is observed (in (30e)).6

Given that scope ambiguity is argued to arise from Quantifier Raising (May 1977, and much work since), the PIPP DO's scopal inertness is either support that the DO is an NP and thus non-quantificational, or that the PIPP DO cannot extract at all for other or related reasons (as seen above), or both. Necessarily narrow scope is a hallmark property of pseudo-incorporated/non-specific objects reported cross-linguistically, again strengthening the argument that PIPP-objects are an instance of such pseudo-incorporation.

To summarize, this first section had two main objectives: the first, to introduce the PIPP construction and to identify a number of interesting restrictions on, and properties of the direct objects that can appear in it. The second was to begin to develop an account of these properties, by highlighting both the similarity of these objects to non-specific pseudo-incorporated objects cross-linguistically, and the syntactic-semantic evidence that these nominals lack D-level structure (which will prove quite important in the next chapters).

6 Non-kind bare plurals have been observed in the literature to show the same scope ambiguities in English as singular indefinites, as in (i):

(i) a. I didn't see parts of that machine. some > every, every > some

(26)

(27)

CHAPTER 2 - THE SYNTAX...

2.1 ...OF (PSEUDO-INCORPORATED) (PSEUDO)PASSIVES: ONE T-VALUE TO SHARE AROUND

To tackle the syntax of pseudo-incorporated pseudo-passives, I must first build up compositionally my assumptions and account of its more basic parts: the passive, then the pseudo-passive. The next section briefly reviews the most basic and relevant aspects of regular passives, and what makes pseudopassives curious, given earlier theories of passive morphology. I argue subsequently that Pesetsky & Torrego's (2004, 2006, 2007; henceforth P&T) framework provides us with a natural way to understand passives, and to reanalyze the old theory of reanalysis (as developed in Hornstein & Weinberg 1981) for pseudopassives, while still capturing the intuitions that incited it. Lastly, I lay out the syntactic problems of fitting an extra direct-object into a pseudo-passive, and how these are circumvented in PIPPs by the very properties we observed in the above section. So now onward.

2.1.1 Passive 101

Since the passive is an integral part of the pseudopassive, and of PIPPs, it is worth going over the core aspects of passivization, which are accepted by and large without much debate. Taking the active sentence in (31 a) as example, it has the related passive sentence in (31 b):

(31) a. David chopped the habanero.

b. The habanero was chopped by David.

Most basically we observe that the subject of (31b) is understood in the same role as the object in (31 a) - in both sentences the habanero ends up chopped, despite that this nominal occupies a different syntactic position in a versus b. In the generative grammar tradition, I take this observation to reflect movement of the habanero from an underlying lower position, where it gets its role-assignment as object, to the subject position in the passive (3 1b). The original agent of the active sentence, "demoted" from its typical subject position, appears optionally in a by-phrase, and the verb itself takes on its past participle formnn while tense is born by the auxiliary be.7

Early theories correlated the past participle morpheme -en with the necessary raising-to-subject of the object, explaining that -en absorbs the verb's accusative case, forcing the object to move to a position where it can receive nominative case licensing (Chomsky 1981, Baker 1988, Baker, Johnson & Roberts 1989). By this analysis, only those verbs that assign accusative case can undergo passivization, to the exclusion of unaccusatives, and unergatives on some accounts. But what about regular pseudopassives? Looking at some examples in (32), we see that they all

There is obviously much more to be discussed and debated regarding the passive, especially with regards to the subject's demotion, how it receives a theta-role, where the by-phrase is positioned syntactically, and the likes. For the purposes of this paper I limit the discussion to the passive properties most relevant to my analysis of PIPPs, but see Baker, Johnson & Roberts (1989), Collins (2005), Hallman (2002), and many others for discussion on the issues I omit.

(28)

fall into the unergative classification, that is they are satisfied with only an agent/subject, though they happily occur with PPs also (pseudopassive forms are primed):

(32) a. Marigold talked (to the students).

a'. The students were talked to by Marigold. b. Ickabod sat (on the big, fluffy couch).

b'. The big, fluffy couch was sat on by Ickabod. c. Each child slept (in a different bathtub).

c'. A different bathtub was slept in by each child.

Pseudopassives are somewhat weird on two counts then: first, it's not clear that the verbs involved assign accusative case to begin with. And second, it's often assumed that nominals within prepositional phrases receive case from the preposition itself, why then should passivization of the verb effect the case-licensing abilities of the preposition embedded below?

We've seen this scenario before, however, namely with long (or super) passives in German, demonstrated in (33):8

(33) a. weil (dieser Turm)i schon vor zehn Jahren ti zu restaurieren versucht wurde. since this tower already from ten years t to restore tried was

"...since _{somebody tried to restore the tower already ten years ago."}

(Wurmbrand, 1998: 147)

In (33) when the verb that embeds the infinitive zu restaurieren 'to restore' is passivized, as versucht wurde 'tried was,' it is the object of the infinitive that raises to subject. That we see a similar phenomenon, passive morphology on one predicate causing a lower predicate to lose its nominal-licensing abilities, with pseudopassives is therefore not so surprising. I propose that such phenomena are in fact easily explained once we adopt the agreement and feature-valuation framework of Pesetsky & Torrego (2004, 2006, 2007) and an extra idea (from Pesetsky p.c.) about passive verbs, which I come to presently.

2.1.2 Pesetsky & Torrego, and Putting the Passive Pieces Together

In Pesetsky & Torrego's (2004, 2006, 2007) framework, Case is an instance of an uninterpretable Tense feature on D, while subject agreement on T reflects the uninterpretable (p-features on T (where p/phi-(p-features are things like number, gender, etc.). Tense needs to agree with a (p-valued D, and a D with valued-Tense, for the former to value its (p-features, and the latter to value its T-features. What is novel in P&T's framework resides in the details of agreement: when two items agree with regards to a given feature, the outcome is two instances of one feature (as opposed to two unique features that have the same value, as in Chomsky 2001), this effectively creates a link between the two feature instances "that is accessible to subsequent processes" (P&T 2007:4). Now when two items agree for a feature that is unvalued on either, the result is not vacuous, but instead yields something like a shared index between the two items, and when one eventually gets valued, the other automatically gets this same value.

8 I _{am grateful to David Pesetsky for initially suggesting this parallel.}

(29)

P&T represent a feature's index within square brackets [ ] (borrowing this notation from the HPSG literature), preceded by a symbol for the given feature ('T' for tense, 'p' for phi/D-features), which is preceded by either a u for 'uninterpretable' or i for 'interpretable.' When one instance of a feature eventually gets valued, its value is indicated just before the bracket (as in uT+past[1]). Therefore an uninterpretable, valued phi-feature on Tense is shown in a tree as uplps[1], whereas an uninterpretable, unvalued (but agreed with) tense feature on D is represented just as uT[ 1] (and whichever head D agreed with for tense would also bear the index [1]). This index sharing mechanism will be important shortly.

A final property of P&T's framework concerns complement selection. The observation is that heads which are needy, i.e. are unvalued for a given feature, select complements that have an instance of this same feature. They call this the vehicle requirement on merge (VRM), as in (34):

(34) Vehicle Requirement on Merge (VRM)

If a and p merge, some feature F of a must probe F on P. (Pesetsky & Torrego 2006:1)

In a canonical sentence (that is, an active one), the verb is lexically valued for tense but unvalued for phi-features, while the noun is lexically valued for phi-features but not for tense. By VRM, for V and D to merge unvalued (p on V must probe for valued cp on D, although this probing results in theta-role assignment instead of agreement.9 Past this point P&T do not elaborate on the whole derivation, but I reconstruct that it would continue as follows. Now TPo,'O which is unvalued for either tense or phi-features, probes into V for valued tense, which V has, so they agree, V raises up to its specifier, and this results in a morphological merger of V into To (in the sense of Matushansky 2002). Since its phi-features are still unvalued, To probes again and finds the DP, who has valued phi-features and unvalued tense to boot. DP is raised to To's specifier (without morphological merger), agrees with T's features, valuing its own uninterpretable tense features as well (which we would otherwise call ACC case licensing). The derivation for a typical transitive sentence like David chopped the habanero looks so far like (35):

(35) TPo DPh T'o To+Vi VP uT[1] *- iT+past[ 1] i(p3s[2] *-- uq[2] th V'

the habanero chopped

ti th

9 P&T (2006) suggest that certain probe-goal relations don't result in agreement because a head must have first doled out all of its semantic relations (theta-roles) before it can value its features.

"0 This is the TensePhrase projected below vP which checks the object's case/T features in a canonical sentence,

much like MP in Johnson (1991), and also known as AgrOP elsewhere.

(30)

Above this, Ts's interpretable tense features must eventually probe into its complement for valued T-features on V+To, then for valued phi-features on the subject DP (which will then raise to specifier of Ts and value its uninterpretable tense features = NOM case):•

TPs DPz

Ts

uT[1] iT[1] i(p3s[3] - u(p[3] David t2 vP S+÷To+Vi TPo... iT +past[1] u(p[2] chopped

Returning now to the English passive, I would like to pick up on an idea from Pesetsky (p.c.) that to be a passive verb is to be unable to lexically bear valued T. Given the canonical derivation above, the effect of merging a tense-unvalued verb is this: V merges with DP just as before, and To with VP. When To agrees with V however, the best they can do is to share an index for tense, since neither has an actual value at this point. When DP is raised, it values the phi-features on V+To, but again gets only a tense-index in return - this DP must, then, raise eventually to specifier of the higher Ts, that is to subject position, where the only valued tense feature of the construction sits, thanks to the auxiliary was.'2 _{Here is the trick: DP finally gets a}

value for tense, but recall that the index that this value replaces is shared by V+To left below also, so via DP's valuation they also get valued."

Following this approach to passives, we get a final tree as in (37) (ignoring vP):

" As mentioned, P&T did not go through a complete derivation such as this in their papers - this is what I think they may have in mind, but it might easily be flawed.

12 Again P&T aren't explicit about what contributes the higher value for T, whether it's v in parallel to V. or

whether it's provided by a tensed auxiliary verb such as was. They also do not specify their assumptions regarding v in passive sentences, that is, if it is defective (they do suggest that To is defective in passives, but I should think this stipulation is no longer necessary once we assume passive V is unvalued for tense). Nothing in my account hinges on one or the other possibility.

13 This is very similar to P&T's (2007) analysis of raising predicates, where the DP that raises to subject values the tense features of the T, v and V projections below in the infinitival complement by agreement with Ts of the matrix verb.

Non-specific objects in the pseudopassive : the syntax and semantics of English pseudo-incorporated pseudopassives

OCT 1

4 2008

L

LIBRARIES

ARCHIVEs

ACKNOWLEDGMENTS

1

Ts